| Summary: | 碩士 === 國立臺灣科技大學 === 電子工程系 === 102 === This thesis investigates the cooperative DSM (CoDSM) technique for future smart grid. In particular, we consider that a load aggregator (e.g., the utility company) coordinates the energy consumption of a neighborhood with a large number of customers, in order to achieve real-time power balance. The deferrable loads (such as the dish washer and washing machine etc.) , adjustable loads (e.g., Electric Vehicles (EV) ) and storage devices (e.g., battery) are considered. Our main interest lies in achieving the real power balance by solving the CoDSM problem in a real-time and fully decentralized manner. To this goal, firstly, we develop a multi-stage linear prediction method for estimating the solar power generation in a short future period of time; secondly, a rolling window based control method, which can exploit both real-time and predicted solar power for real-time CoDSM is used; thirdly, distributed optimization methods are applied. Specifically, we study the load scheduling performance of three state-of-the-art distributed algorithms, namely, the distributed dual subgradient (DS) method, the dual consensus subgradient (DCS) method and the dual consensus alternating direction method of multipliers (DC-ADMM). Simulation results show that the distributed CoDSM algorithms can improve the power balance significantly and the DC-ADMM method performs best.
|
|---|
